Twitter reveals NYC neighborhoods' racism and homophobia, researchers find

Twitter may hold a key to pinpointing precise neighborhoods where systemic racism and homophobia have taken hold, allowing researchers to measure the mental-health impact on diverse groups of people who spend time in those places.

That's what a team from NYU Tandon School of Engineering advance in a pioneering new study published this month in Social Science & Medicine.

Rumi Chunara—an associate professor in NYU Tandon's Computer Science department and in NYU School of Public Health's Biostatistics department—and her colleagues divided up all of New York City into more than two hundred geographic clusters—discrete areas, in some cases just multiple blocks—defined by the degree of racism and homophobia within each of those locations.

Unlike typical analysis of structural discrimination that relies on public records like census or housing data, Chunara and her colleagues used Twitter as a proxy to measure nuanced "on the ground" negative sentiment that can permeate the cultural climate of specific locations.

The complex analysis—conducted originally in 2018 and published in Proceedings of the ACM on Human-Computer Interaction—required the team develop and employ a novel technique called socio-spatial self-organizing maps (SS-SOMs), in which they applied a combination of keyword filtering, labeling, and iterative learning to Twitter posts generated in New York City. With the resulting data, they drew boundaries around small areas based on the similar racial and sexual-orientation attitudes expressed in the postings.

For this project, the team built on that existing research by recruiting 147 young Black, white and Hispanic men in New York City who identify as gay or bisexual, to wear GPS tracking devices for a two-week period. The participants self-reported their mental health during that time, accounting for feelings like stress or depression.

All three data sources—Twitter sentiment, GPS tracking and self-reported mental health status—gave researchers the ingredients for a regression analysis that investigated how spending time in places with negative attitudes about race or sexual orientation affected the men's mental well-being. Researchers also looked at whether the participant's race or ethnicity affected this relationship.

"Structural racism and homophobia have been uniquely challenging to measure scientifically, and thus understanding their effects on health has been difficult," said Chunara. "We knew from prior research that social media, and Twitter specifically, provides a valid proxy for 'on-the-ground' attitudes of a given location, a component of structural discrimination that isn't always visible in data from traditional public records. This type of Twitter data provides a foundation on which we can then research nuanced questions about the impact of placed-based negative sentiment in people's lives. The breakthrough in this research is showing how the 'on the ground' attitude affects mental health for men in sexual and racial minorities. Access to Twitter is critical for this type of research, that ultimately gives us information that can be used to improve mental health for diverse groups of people."

The study concludes that a complex relationship seems to exist between the time participants spent in areas with higher levels of racist and homophobic sentiment and their mental well-being. Researchers found noticeable differences between Black and white non-Hispanic men and Hispanic men, with Black and white men reporting significantly more poor mental health days compared to Hispanic men when exposed to high levels of negative racial attitudes.

"This research shows that it's possible to measure the ways that a climate of racism and homophobia impacts the lived experiences of people with different racial and sexual-orientation identities," said Chunara. "It lays the foundation for future research that can use novel tools like SS-SOMs that we've tested and fine-tuned in this project, to explore these issues more deeply in bigger studies."